The contingent negative variation (CNV) is an event-related potential (ERP) indicative of cognitive processes in continuous performance tasks (CPTs; Bekker et al., 2004). First described by Walter et al. (1964), the CNV is a slow, negative-polarity ERP occurring between a warning stimulus (S1) and an imperative stimulus (S2) (Tecce, 1972; Brunia and van Boxtel, 2001; Ahmadian et al., 2013). The CNV amplitude increases in anticipation of task performance and is thought to index changes in neural excitability in preparation for an approaching internal or external stimulus (Kononowicz and Penney, 2016). Furthermore, the CNV is associated with motor preparation and linked to response time (RT) and response time variability (Karalunas et al., 2014). The CNV is widely distributed over the scalp, with amplitudes typically peaking at frontal and central electrodes (Hamano et al., 1997). The primary neural generators of the CNV are thought to be located in the frontal cortex (Oishi and Mochizuki, 1998; Segalowitz and Davies, 2004; Battaglini et al., 2017), which plays a central role for exerting top-down response preparation and control (Rosahl and Knight, 1995; Brunia, 1999; Stuss and Alexander, 2007).

The CNV has been investigated mainly within the interpretative framework of executive functions (Flores et al., 2009). Executive functions represent higher-order cognitive processes that support goal-oriented behaviors, such as inhibitory control, problem solving, and cognitive flexibility (Cristofori et al., 2019). Imaging studies have demonstrated that these functions are represented by widespread brain networks, mainly localized in the pre-frontal cortex in conjunction with subcortical and allocortical regions, such as limbic areas and the hippocampus (Cristofori et al., 2019).

In typically developing (TD) participants, executive functions evolve at a different pace throughout development and continue to mature in adolescence (Best and Miller, 2010). Despite some older studies reporting an increasing CNV in pre-adolescence with maximum amplitude at 15 years (Tecce, 1971, 1972), more recent studies suggest that the CNV amplitude increases linearly from childhood through adolescence and adulthood, in line with the assumed maturation of executive function abilities in general (Segalowitz and Davies, 2004; Jonkman, 2006; Segalowitz et al., 2010). In TD children and adults, a larger CNV amplitude is associated with stronger response preparation and better performance in executive function tasks (Segalowitz et al., 1992; Dywan et al., 1994; Jonkman, 2006).

Another interesting research avenue has sought to unravel group differences between TD controls and psychiatric patients regarding the CNV as an indicator of executive processes. For instance, in several neuropsychiatric disorders, a lower CNV amplitude was found compared to TD controls, such as patients with schizophrenia and bipolar disorder (Li et al., 2015), patients with schizophrenia (Wang et al., 2020) and patients suffering from seizures (Drake et al., 1997). The CNV has also been investigated as a neurophysiological biomarker in functional movement disorders (Teodoro et al., 2020), psychomotor dysfunction in schizophrenia (Juston Osborne et al., 2020) and post-traumatic stress disorder after an acute trauma (Kowalski et al., 2018). Moreover, the CNV has also been utilized to asses changes in the dopaminergic function induced by pharmacological treatments (Linssen et al., 2011). However, besides a study on the CNV as a treatment response to an intervention targeting abnormal attention in functional movement disorders (Teodoro et al., 2020), almost all studies investigating CNV changes as an outcome measure were conducted as part of ADHD neurofeedback research (for an overview see: Mayer et al., 2013). These intervention studies showed increases in CNV amplitudes and reduced ADHD symptomatology after neurofeedback training (Mayer et al., 2013, 2016; Gevensleben et al., 2014).

Another common psychiatric disorder associated with an array of executive dysfunctions is Autism Spectrum Disorder (Demetriou et al., 2018). Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder defined by social and communicative deficits and repetitive, stereotyped behaviors (International Classification of Diseases ICD-10; WHO, 2016). Impaired executive functions in ASD have been suggested as a model for understanding behavioral problems associated with the disorder (Demetriou et al., 2018, 2019). Furthermore, common impairments in the executive function abilities such as set shifting, response inhibition, and working memory negatively impact indicators of mental health, social interaction, and lifelong functioning outcomes in ASD (Demetriou et al., 2019).

Neuroimaging studies have linked difficulties in executive functions to altered functional connectivity in ASD (Kana et al., 2007; Koshino et al., 2008; Braden et al., 2017). A recent meta-analysis reported ASD-related activation abnormalities in the left inferior frontal gyrus, left inferior parietal gyrus and sensorimotor areas during the performance of executive function tasks (Zhang et al., 2020). The results suggest that the mechanisms underlying executive function difficulties in ASD are relatively stable across all age groups, which is in line with behavioral data suggesting executive dysfunctions throughout development (Lai et al., 2017; Zhang et al., 2020). However, despite measurable executive dysfunctions, adults with ASD show a slightly better performance on some executive function tasks due to improved compensatory strategies and/or developmental maturity (Demetriou et al., 2018). Thus, investigating the CNV in ASD populations might provide valuable information beyond behavioral data about the underlying neurocognitive sources of executive difficulties.

However, studies analyzing the CNV in participants with ASD have yielded mixed results (Strandburg et al., 1993; Tye et al., 2014; Thillay et al., 2016; Høyland et al., 2017; Hoofs et al., 2018). Early studies could not find any differences between CPT performance and CNV amplitude of high-functioning adults with ASD compared to TD adults (Strandburg et al., 1993). This result was partly supported by a study on adolescents with ASD, showing no differences between adolescents with ASD compared to TD controls aged between 12 and 15 years, whereas for older adolescents (older than 16 years), the CNV was found to be enhanced (Høyland et al., 2017). Similarly, Tye et al. (2014) reported increased CNV amplitudes in participants with ASD compared to participants with ADHD, ASD + ADHD, or TD controls. No differences in the behavioral outcomes of the CPT were registered between children with ASD and TD children (Tye et al., 2014). Thus, comorbidity factors seem to affect the expression of CNV in ASD and need to be taken into account in future analyses. In addition, Thillay et al. (2016) found an increased CNV amplitude comparing adults with ASD and TD adults. Only one study reported a decreased CNV in adolescents and young adults with ASD compared to a TD control group (Hoofs et al., 2018). In this study, the ASD group did not show CNV differentiation based on the task specifics, indicating that ASD participants may have deficits in fine-tuning their intentional control to negligible conditions (Hoofs et al., 2018). This result further points toward an altered top-down response preparation in ASD (Høyland et al., 2017).

With this study, we aim to quantify the effects of neurofeedback training on the CNV in children and adolescents diagnosed with ASD. Given the mixed and scant results obtained in the context of ASD and CNV, we refrained from formulating a directional hypothesis and performed our analyses in an exploratory manner. Furthermore, as indicated in previous studies, we explore the role of three components of comorbid ADHD symptoms, namely, inattention, hyperactivity, and impulsivity, as potential moderators. In addition, we consider the age of participants as a potential control variable, as suggested by the literature. In order to utilize the entire data available, we fit a multivariate hierarchical Bayesian model to trial-level CNV amplitudes obtained prior to and following neurofeedback training. Finally, we elucidate the importance of using the finer-grained separation of comorbid ADHD symptoms instead of aggregate comorbid scores.

In this work, we investigated changes in the CNV following an intensive neurofeedback training for adolescents with ASD and comorbid ADHD symptoms. In order to obtain a more nuanced exploratory analysis, we formulated and fitted two hierarchical moderator models: one model containing a binary aggregate ADHD score as a single moderator and one model containing three separate moderators representing quasi-continuous scores for inattention, hyperactivity, and impulsivity. Our main goals were to show that (i) utilizing the non-aggregated trial-based EEG data can yield meaningful predictions and uncertainty intervals; (ii) comorbid ADHD symptoms comprise an important factor to consider when investigating the CNV in ASD populations and (iii) disentangling binary ADHD scores into their underlying sub-components can provide more information on how comorbidity in ASD affects changes in CNV.

Accordingly, our binary moderator model reveals no influence of the aggregate ADHD score on the observed CNV trajectories (slopes) across the two time points. However, differentiating between adolescents with comorbid ADHD and adolescents without comorbid ADHD, the model predicts radically different CNV amplitudes between the control and the neurofeedback group at baseline. Participants with ASD and comorbid ADHD exhibit similar CNV amplitudes prior to the intervention, irrespective of their group. After the intervention, the average CNV amplitude in the neurofeedback group increases, whereas the average CNV amplitude in the control group decreases, resulting in a diverging pattern (cf. Figure 3, left panel). Differently, participants without comorbid ADHD exhibit widely different CNV amplitudes prior to the intervention: The average CNV amplitude in the control group is much higher than the average CNV amplitude in the neurofeedback group. Following the intervention, the average CNV amplitude in the neurofeedback group increases, whereas the average CNV amplitude in the control group decreases, resulting in a converging pattern (cf. Figure 3, right panel). The results obtained for participants without comorbid ADHD are difficult to interpret due to the differences in baseline CNV, since treatment-induced change might easily be confounded with regression to the mean. Larger samples or hypothesis-driven stratification for small samples might overcome unbalanced baseline distributions in future studies.

The fine-grained moderator model implies a rather complex pattern. In the control group, the CNV amplitude either decreased or remained relatively stable across the eight symptom-severity combinations. In the experimental group, two different patterns emerged depending on the severity of comorbid impulsivity symptoms. Whereas participants with low impulsivity scores showed a decreased CNV amplitude, participants with high impulsivity scores exhibited an enhanced CNV amplitude after neurofeedback intervention (cf. Figure 4, second row). The other two comorbid symptoms – inattention and hyperactivity – did not seem to exert a noticeable influence on the group-dependent change in CNV amplitude. This suggests that impulsivity might be the primary ADHD component driving changes in CNV. Considering the current results, future studies investigating the response of the CNV to different clinical interventions in ASD should include a measure of impulsivity as a moderating factor.

Furthermore, age appears to be an important predictor of CNV in the fine-grained model, implying that the average CNV amplitude increases with participant’s age. Considering that the CNV changes with age in TD persons (Jonkman, 2006), this result is line with previous research. However, results regarding CNV amplitudes and developmental trajectories in ASD are mixed (Strandburg et al., 1993; Tye et al., 2014; Thillay et al., 2016; Høyland et al., 2017) and the current study only provides a cross-sectional view on the effects of age.

Although impulsivity as a construct is generally well investigated, a comprehensive theory which overcomes major challenges of theoretical specifications, multiplicity of definitions, and measurement approaches, is still missing (Strickland and Johnson, 2021). Moeller et al. (2001) described impulsivity as a biopsychosocial construct encompassing a reduced sensitivity to negative consequences of behavior, rash responses to stimuli without thoroughly processing the available information, and little to no regard for long-term consequences of behavior regarding all actors involved in a situation. Further characteristics of impulsivity are a lack of inhibition and reflection in decision making, deficits in delaying gratification, premature behavior and inability to resist impulses and urges (Robbins et al., 2012; Berlin and Hollander, 2014). Impulsive behavior is commonly described in diverse psychiatric disorders, such as ADHD, addiction, or bipolar disorder as well as in antisocial and aggressive disorders (Moeller et al., 2001; Robbins et al., 2012). Furthermore, impulsive behaviors are considered as a risk factor for the development of substance abuse and addiction (Verdejo-García et al., 2008; De Wit, 2009; Robbins et al., 2012) and facilitate non-suicidal self-injury behaviors and suicide attempts (Corruble et al., 1999).

Frontal brain areas have been reported as common neural substrate for impulsive behaviors (Berlin and Hollander, 2014). In particular, fronto-striatal circuitries, stemming from striatal projections, have been associated with impulsive behavior and deficits in inhibitory prefrontal control (Dalley et al., 2011; Berlin and Hollander, 2014). Also, studies have found an association between the generation and amplitude of the CNV and the activation of a thalamo-cortico-striatal network (Ikeda et al., 1997; Nagai et al., 2004; Fan et al., 2007). BOLD activity has been associated with the trial-by-trial variation in CNV amplitude in the bilateral thalamus, the anterior cingulate, and the supplementary motor cortex (Nagai et al., 2004). Strong relationships between the CNV and impulsivity have also been reported in offenders with antisocial disorder, for whom the CNV have already been suggested as a potential predictor of recidivism (Howard, 2002). Therefore, impulsivity has been assumed to arise as a result of impaired response inhibition, that is, a deficient top-down control (Dalley et al., 2011).

Neurofeedback training has been suggested as a viable treatment for impulsivity and inattention in children with ADHD (Arns et al., 2009). However, in our fine-grained moderator model, inattention seemed to exert little influence on CNV patterns and we could not find intervention-specific behavioral improvements in the CPT regarding sustained attention, which is in line with CPT results by Pamplona et al. (2020). Nevertheless, using network-based real-time functional magnetic resonance imaging, Pamplona et al. (2020) found improved temporal sustained attention following neurofeedback training. Thus, initial results underline the potential of neurofeedback as a non-pharmacological method to enhance attention and possibly other cognitive functions in psychiatric disorders.

Regarding further neurophysiological changes, increased CNV amplitudes have been found in patients with ADHD after neurofeedback training (see Mayer et al., 2013). However, to our knowledge, no other study explored the effects of subcomponents of ADHD symptomatology on CNV changes. Importantly, ADHD is a common comorbidity of ASD, with co-occurrence rates ranging from 37% to 85% (Leitner, 2014; Stevens et al., 2016; Gnanavel et al., 2019). Both disorders have been associated with abnormal allocation of attentional resources and performance monitoring (Lau-Zhu et al., 2019).

A consideration of ADHD and ASD from a transdiagnostic perspective based on the Research Domain Criteria Initiative identifies three distinct transdiagnostic subtypes of executive function profiles. Subtypes are defined by deficits in (1) behavioral flexibility and emotion regulation; (2) hyperactivity/impulsivity and inhibition; and (3) working memory, organizing, and planning (Vaidya et al., 2020). Alongside this subtype focus, our findings contribute to the investigation of potential alternative or adjunct treatment approaches for ASD, revealing fine-grained differences in expected treatment outcomes depending on the comorbid subtypes or even a transdiagnostic psychopathological construct.

In summary, our study demonstrates that improvements in executive functions after SCP neurofeedback training, as indicated by the increase in the CNV following intervention, is limited to the group of young patients with deficits in impulsivity control. Future research is necessary to investigate whether the transdiagnostic subtypes suggested by the Research Domain Criteria Initiative are linked to different functional and neuroanatomical networks. In this way, impulsivity would not only be recognized as an underlying construct for several disorders in ICD or DSM (Moeller et al., 2001), but also its transdiagnostic neuropsychological mechanisms and subtle impact across mental disorders would be put in the spotlight.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Ethics Committee of the Medical University of Vienna. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

LK and LP conceived and designed the study. KP, SR, and KD conducted the measurements. SR, KP, and LK analyzed and interpreted the data. KP, SR, KD, LP and LK wrote, edited, and revised the manuscript. All authors contributed to the preparation of the manuscript and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.